OpenAI says it is cutting the price of GPT-5.6 Luna by ~80% and the price of GPT-5.6 Terra by 20% after improving the efficiency of the systems that serve them
THE SO WHAT
An ~80% price cut on Luna and 20% on Terra compresses the cost of “good enough” intelligence—many SaaS AI markups just got harder to defend. Teams should revisit unit economics on any product reselling OpenAI calls and consider moving more workloads from in-house models to API if they can’t match this curve.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIGoogle says it fixed more Chrome bugs in June than over the past two years, thanks to AI
Bug discovery is becoming a compute problem, not a headcount problem—LLM-assisted fuzzing and code analysis just pulled two years of Chrome fixes into a single month. If you ship complex software and aren’t running AI across your codebase and crash logs, assume your defect backlog is now a competitive liability.
Applied AILinkedIn actually adds a ‘seems like AI slop’ button
A dedicated “seems like AI slop” flag means platforms are starting to algorithmically downrank low-effort AI content—and users are being deputized to help. If your marketing or recruiting teams are flooding LinkedIn with generic AI-written posts, expect diminishing reach and start investing in signal, not volume.
Applied AIDeepSeek Is Developing Massive AI Data Center in Inner Mongolia
A “massive” AI data center in Inner Mongolia is about cheap land and power—China is pushing frontier-scale training into interior regions the way US hyperscalers went to West Texas and Oregon. If you depend on Chinese models or compete with them, factor in that their effective compute ceiling is rising via domestic infra, not just imported GPUs.
Applied AIAt a hearing, a US judge says "I don't see additional evidence" from the Pentagon justifying its designation of Anthropic as a supply-chain risk
A federal judge openly questioning the evidence behind a supply-chain risk designation shows how fluid AI vendor blacklists still are. If you sell into government, don’t assume current “trusted” or “banned” lists are stable—build scenarios where procurement rules swing quickly in either direction.